ollama - 💡(How to fix) Fix Total bug [5 comments, 3 participants]

TL;DR

• Downgrade Ollama to a previous version may resolve the high CPU utilization issue with certain models.

Guidance

• Investigate the changes made in Ollama version 0.20.3 to identify potential causes of the increased CPU load.
• Test the same models with a smaller context window to see if the issue is related to the token size.
• Verify if other users have reported similar issues with the same version and models to confirm if it's a widespread problem.
• Consider reaching out to the Ollama development team for guidance on resolving the issue or providing a patch.

Notes

• The issue seems to be specific to Ollama version 0.20.3 and certain models, so the solution may not be applicable to other versions or models.
• Without more information about the changes made in version 0.20.3, it's difficult to provide a more specific fix.

Recommendation

• Apply workaround: Downgrade to a previous version of Ollama, as it is reported to have worked smoothly with the mentioned models.